Because modern telecommunication systems require extremely stable and accurate timing devices, atomic frequency standards have been used in such applications. For this and other applications, the overall size, operating temperature, power consumption, weight and ruggedness of the atomic standard are critical parameters.
Microwave atomic frequency standards use natural resonances within atoms to keep time since the natural atomic resonances are more stable and less sensitive to environmental effects, such as temperature, pressure, humidity, vibration, acceleration, etc., than are macroscopic oscillators like pendulums and quartz crystals. The natural atomic resonances are generally ground-state microwave hyperfine transition frequencies of the atoms in an atomic gas. This type of quantum atomic oscillator, operating at the hyperfine transition of an atomic gas, is used as a highly stable frequency reference to which the frequency of a variable frequency oscillator, such as a quartz oscillator, can be electronically locked so the high stability and relative insensitivity to environmental perturbations of such a natural atomic resonance are transferred to the quartz oscillator.
Atomic frequency standards usually comprise an electronic assembly including a voltage-controlled oscillator (VCO), and a physics package that maintains an accurate and stable VCO standard frequency on a long-term basis. The physics package and associated electronics are used to slave the VCO output to the selected hyperfine transition frequency of the quantum system, to thereby reduce frequency drift due to oscillator aging and the effects of the environment on the oscillator.
In atomic frequency standards in which the quantum system comprises a passive gas cell containing an atomic gas, such as rubidium or cesium, the physics package includes a light source, such as a plasma discharge light source or a semiconductor laser, a transparent gas cell (the resonance cell) and a photo detector for light that is transmitted through the atomic gas. The resonance cell, and sometimes the photo detector, is located in a microwave cavity, which is resonant at the hyperfine transition frequency of the atomic gas. The resonance of the microwave cavity at the hyperfine transition frequency of the atomic gas is used to maximize the effect of an injected microwave electromagnetic field on the atomic gas in the cell. The injected microwave electromagnetic field is generated by frequency multiplication and synthesis from the VCO output. The generated microwave frequency, that is approximately equal to the hyperfine transition frequency of the atomic gas in the cell, is then modulated (frequency modulation) and the microwave energy is injected into the microwave cavity.
If the atomic gas is an alkali vapor (e.g., Rb or Cs), the light that is generated by the light source includes light at one or both of the D1 and D2 optical atomic resonance frequencies (such light frequencies being referred to hereinafter as “D transition light”). (D1 transition light is resonant with the optical transition between the 2S1/2 state and the first 2P1/2 optically excited state of the alkali atom, while D2 transition light is resonant with the optical transition between the 2S1/2 ground state and the first 2P3/2 optically excited state.) In the case of rubidium (Rb), the D1 transition light and D2 transition light have wavelengths (optical frequencies) of 794.8 nm (377 THz) and 780.0 nm (384 THz), respectively. In the case of cesium (Cs), the wavelengths (optical frequencies) of the D1 transition light and D2 transition light are 894.6 nm (335 THz) and 852.3 nm (352 THz), respectively. (1 THz=1012 Hz).
In operation, an alkali vapor, such as Rb or Cs, within the resonance cell in the microwave cavity is optically pumped by D transition light from the light source. In the absence of optical pumping, the populations of the two ground-state hyperfine levels are nearly equal. The effect of the optical pumping is to create a population difference between these levels; and, in the process of optical pumping, light is absorbed by the atoms. The microwave energy, injected into the microwave cavity at about the hyperfine transition frequency of the atomic gas, interacts with the atoms of the atomic gas that are in the ground-state hyperfine levels, inducing transitions between these levels and tending to drive the population difference to zero (i.e., equalize the populations). The optical pumping process, on the other hand, tends to maintain this population difference by the optical absorption of the D transition light. The optical pumping proceeds at the maximum rate when the difference between the injected microwave frequency and the hyperfine transition frequency of the atomic gas is zero and, as a result, the light energy absorbed by the atomic gas is maximized, and the light intensity sensed by the resonance cell photo detector is minimized. The intensity of the light that is transmitted through the atomic gas in the gas cell is sensed by the photo detector and the variation in light intensity is detected by the photo detector and used to generate a control output that locks the VCO output to the stable hyperfine transition frequency of the atomic gas.
It is well known that an alkali vapor, plasma discharge light source produces both D1 and D2 transition light that contains optical hyperfine components which must be removed or reduced in order to increase the efficiency of optical pumping to a practically useful value. These closely spaced, unwanted hyperfine components cannot be easily removed using conventional optical filtering; but, in the case of Rb vapor, there is a simple solution: isotopic filtering is employed instead using either (1) a separate, temperature-controlled isotopic filter cell containing 85Rb that is placed between the light source and the resonance cell, or (2) a resonance cell that combines the resonance and filter functions by adding 85Rb to the 87Rb already in the resonance cell (an integrated cell, so-called because it integrates the two functions). It should be noted that isotopic filtering removes part of the D1 transition light and D2 transition light, but some of the D1 and D2 transition light is still present after isotopic filtering and both perform the optical pumping function simultaneously.
If a semiconductor laser is used as the light source, it must operate in a single, longitudinal, transverse and polarization mode at an optical frequency (wavelength) that can be tuned to either the D1 transition light wavelength or the D2 transition wavelength. In this case, there are no unwanted optical hyperfine components because the laser output is a single, well-defined optical frequency (wavelength), and no optical filtering of any kind is needed.
In most applications, atomic frequency standards are used over a range of ambient temperatures, with the requirement that the temperature sensitivity (the change in the output frequency of the standard as a function of ambient temperature) be very small. Achieving this result is complicated by the fact that there are many potential contributions to the temperature sensitivity (TS) from both the electronics and the physics package of the device. (See, for example, W. J. Riley, “The Physics of the Environmental Sensitivity of Rubidium Gas Cell Atomic Frequency Standards,” IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, Vol. 39, March, 1992, pp. 232-240).
In the case of optically pumped gas-cell atomic frequency standards, there are several mechanisms that can cause TS to change with time. One of these mechanisms is a result of the combined action of the light shift, the change of light intensity with ambient temperature, and the change over time of the light intensity from the light source.
The change in TS due to this mechanism can be expressed approximately in the following simplified way for an atomic frequency standard using an integrated resonance cell. Let I=light intensity entering the resonance cell, y=fractional frequency offset from the standard output frequency f=Δf/f and yLS=the change in y due to the light shift, then the two are related by a light-shift coefficient kLS according to the equation,yLS=kLS·IyLS(TC, TL, T, t)=kLS (TC, TL, T)·I(TL, T, t),where TC=cell-oven set-point temperature, TL=lamp-oven set-point temperature, T=ambient temperature, and t=time. The dependence on ambient temperature results from: (1) the finite loop gain of the thermostats controlling the lamp and cell oven temperatures, and (2) the temperature gradients across the lamp and cell. (The crucial temperatures are those at the locations of the alkali-metal deposits in the lamp and cell, but these temperatures change with ambient temperature due to: (1) departures from the set-point values, and (2) the temperature gradients over the lamp and cell that change with ambient temperature and the fact that an alkali-metal deposit is at a different location than its corresponding oven-temperature sensor.)
During normal device operation, the set-point temperatures are not changed; i.e., TL and TC are parameters that are held constant (but the temperatures of alkali metal deposits in the resonance cell are not constant, but change with ambient temperature). In this case, the above equation can be written in simplified form as,yLS(T, t)=kLS(T)·I(T, t).If the ambient temperature changes from T1 to T2 over a time period during which the light intensity remains sensibly constant, the contribution to TS from the light shift is,TSLS(t)=yLS(T2, t)−yLS(T1, t).Next, consider what happens if after a (long) time period, Δt, there has been a change in the light intensity due, for example, to light intensity decay from the light source. The contribution to TS from the light shift in this case is,TSLS(t+Δt)=yLS(T2, t+Δt)−yLS(T1, t+Δt).The change in TS due to the light decay is then,                               Δ          ⁢                                           ⁢                      TS            LS                          =                ⁢                                            TS              LS                        ⁡                          (                              t                +                                  Δ                  ⁢                                                                           ⁢                  t                                            )                                -                                    TS              LS                        ⁡                          (              t              )                                                              =                ⁢                                                            k                LS                            ⁡                              (                                  T                  2                                )                                      ·                          [                                                I                  ⁡                                      (                                                                  T                        2                                            ,                                              t                        +                                                  Δ                          ⁢                                                                                                           ⁢                          t                                                                                      )                                                  -                                  I                  ⁡                                      (                                                                  T                        2                                            ,                      t                                        )                                                              ]                                -                                                ⁢                                                            k                LS                            ⁡                              (                                  T                  1                                )                                      ·                          [                                                I                  ⁡                                      (                                                                  T                        1                                            ,                                              t                        +                                                  Δ                          ⁢                                                                                                           ⁢                          t                                                                                      )                                                  -                                  I                  ⁡                                      (                                                                  T                        1                                            ,                      t                                        )                                                              ]                                ,                    which is generally non-zero since the light shift coefficient at T1 ambient is not the same as the light shift coefficient at T2 ambient and the light intensity is not constant over long periods of time.
It follows from the last equation that if I(T, t+Δt)=I(T, t) at all values of T, then ΔTSLS=0, even though the light intensity is still allowed to vary with ambient temperature. To put it another way, if I0(T)=I(T, t) is known and stored, and if at all later times t′, the table I(T, t′) can be made equal to the original table, I(T, t′)=I0(T), then the TS due to the light shift will not change with time. To put it still another way, if the light intensity of the light source varies with ambient temperature, I=I(T), then this will generally result in a non-zero contribution to the temperature sensitivity of the standard due to the light shift effect; i.e., TSLS≠0. As long as I(T) doesn't change with time, then whatever the value of TSLS is, it will stay the same; i.e., the only way the temperature sensitivity due to the light shift can change with time is if the light intensity itself changes with time.
(For an atomic standard using a separate isotopic filter cell instead of an integrated cell, it can be seen that, (1) the above equations still apply if the argument “TC” is replaced by the arguments “TR, TF” where TR=non-integrated resonance cell set-point temperature and TF=filter cell set-point temperature, and (2) the above conclusions are unchanged.)
Stabilization of the light intensity from electrodeless discharge lamps has been taught in the prior art, a brief description of which follows.
U.S. Pat. No. 2,975,330 (Bloom & Bell, 1961) describes the use of a light pipe and a photocell to detect light from an alkali-metal electrodeless discharge lamp for the purpose of stabilizing the light intensity. The signal from the photocell is electronically processed and fed back to control the output power of the rf oscillator producing the discharge.
U.S. Pat. No. 4,431,947 (Ferris & Shernoff, 1984) describes the intensity stabilization of 254 nm (ultraviolet) light intensity from a mercury electrodeless discharge lamp, an arrangement similar to that of Bloom & Bell, but with the addition of an optical bandpass filter. The optical filter allows only 578 nm to pass so the intensity of 578 nm (green) light is directly stabilized whereas additionally, due to the properties of the photo detector and the discharge, the 254 nm light intensity is indirectly stabilized.
In both of the above patents, the light from an electrodeless discharge lamp was stabilized by directly sensing it using a photo detector whose electrical output was then used to control the power output of the rf oscillator that produced the discharge. Neither of these older methods is presently satisfactory because they require an additional photo detector (for sensing the light directly at the output of the light source).
In today's subminiature gas-cell standards, it is more difficult to control the temperature of the light source and the gas cell because the small size of the standard tends to reduce the amount of thermal insulation between the light source and gas cell and the outside ambient temperature. If this causes the light source to experience significant temperature change when the ambient temperature changes, then the light-shift contribution to TS, and ΔTS as well, will not be negligible.
The purpose of this invention is to greatly reduce the size of ΔTS by stabilizing the light intensity of the light source in the presence of aging and varying ambient temperature, thereby making I(T, t+Δt)−I(T, t)≈0 at all values of ambient temperature. While the effect of this improvement is to make ΔTS≈0, there is still a light shift contribution to TS, but now this contribution does not change significantly with time.